[Ntp] Antw: Re: Antw: Re: Antw: Re: Calls for Adoption -- NTP Extension Field drafts -- Four separate drafts

["Ulrich Windl" <[e-mail]>]

>>> Hal Murray <[e-mail]> schrieb am 02.09.2019 um 11:37 in
Nachricht <[e-mail]>:

> [e-mail]‑regensburg.de said:
>> But remember that a "node" is NOT "an IP address". With multi‑homed hosts
you
>> can have interesting synchronization loops... ;‑) 
> 
> Is there a problem with using an IPv6 address to identify a system?  (if you

> are multi‑homed, just pick one)
> 
> You have to skip over fe80::, Link‑Scoped Unicast.  Are there other ranges?

Once we tried anycast (see NTP bug 3366). It was IPv4 only (eccept maybe
link-local IPv6 addresses assigned automatically)

> 
> There is some way to map IPv4 addresses to IPv6 if you aren't using IPv6.  
> RFC‑1918 addresses won't be globally unique.  They should work if you are on

> an isolated network.  That doesn't work if you have more than one system 
> going through the same NAT box.
> 
> Would it be simpler/cleaner to use an Ethernet host addresses?  (what are 
> they called these days?)  Are there any potential cases where an NTP server

> wouldn't have one?

I think the only reasonable solution would be "server ID" that covers all
possible network adresses of one server. Whether that server ID is just a
random value or maybe a hash of the FQHN is an open question, but IMHO 64 bits
should be enough (128 seems overkill, and 96 doesn not make much sense IMHO).

Regards,
Ulrich

> 
> 
>  
> 
> ‑‑ 
> These are my opinions.  I hate spam.
> 
> 
> 
> _______________________________________________
> ntp mailing list
> [e-mail] 
> https://www.ietf.org/mailman/listinfo/ntp